QRE adds errors to human decisions, keeping the consistency of beliefs and decisions. The
stochastic generalization via a logistic choice rule of the Nash equilibrium has been shown
to have two appealing features: (i) its comparative statics properties are in some cases more intuitive than for the standard NE, and (ii) it explains experimental data that are consistent
with intuition but are not predicted by a Nash equilibrium. In this paper we show that level k
model, much simpler in terms of rationality requirements, reproduces the same
comparative statics and offers predictions close to the logit QRE within the class of games
with the local payoff property